Folding Forum

An RTX 3090 on Linux earns around 9M PPD. By utilizing a trick called MPS (Multi-Process Service), this can be increased to 14M PPD. Other big GPUs, like the 4080/4090 and 5080/5090, may see even greater gains. Remember that smaller GPUs do not benefit from this guide. If you already know a bit abou...

The reason the WU upload results doesn't progress is a known bug: https://github.com/FoldingAtHome/fah-client-bastet/issues/355 The reason the WU is marked in history as failed is another known bug: https://github.com/FoldingAtHome/fah-client-bastet/issues/352 In spite of the history marking it as f...

How valuable is 3D V-Cache vs higher clock rates then? I'm looking for an EPYC but there are a whole lot of options. I'm considering either a 9375F (4.4 GHz all core boost, 32 MB per CCD = 256 MB L3 total, 6400 MT/s memory) or a 9384X (3.5 GHz all core boost, 96 MB per CCD = 768 MB L3 total, 4800 MT...

It is getting 100% workload because GPU folding on Nvidia runs a single core at 100% to "feed" the GPU. So only three cores are folding the CPU task and one of the three cores is feeding the GPU, but it adds up to all four cores running at 100%.

"Retry 217" tells me that the client has been running for quite some time. Try restarting the service. My guess is it will work. I seem to recall an issue where changes to resources while stuck in "assign wait" don't always take effect until a restart of the client.

miziqaz This kind of answer IS OF NO HELP at all. The issues you are seeing appear to be related to stability. This isn't something we're able to help with. The computer is probably faulty, but you only found out when trying FAH. A faulty (unstable) computer might be able to browse the web or play ...

Nvidia has a guide on how to use MPS with OpenMM now: https://developer.nvidia.com/blog/maxim ... s-service/

If it's hanging at "DOWNLOAD 100%" then it's most likely this known bug: https://github.com/FoldingAtHome/fah-cl ... issues/355

Yeah they are very fine grained. I think they will stop at most 25 steps past the interrupt signal, which takes no more than a second or two, and they resume from exactly the step that it stopped.

First check the beta forum by project number and see if it has been listed as moved to FAH. Second, the usual summary list only shows those projects which are on FAH. Beta are listed separately the last I checked. There is a separate Project Summary for Beta, I would have to look up the link. Maybe...

For GPU cores, this is tracked in issue #247 . It may be a while before anything like that is implemented for GPU cores. For CPU cores, it would be quite easy but the performance improvement would be negligible because the checkpoints are very fast anyway. The best way to speed up CPU checkpoints wo...

Maybe it doesn't support all the ROCm features, but I can use rocm-smi and rocminfo on my 780M at least. It lacks many features that would be useful for professional use like setting a power cap or getting PCIe bandwidth usage, but it mostly works. $ rocm-smi ======================= ROCm System Mana...

There is the latest https://github.com/foldingathome/fah-client-bastet that you can compile yourself which has newer code than the 8.4.9 tag, but it is untested and doesn't contain any code changes that would be relevant to the problem you're having anyway.

This will be addressed by https://github.com/FoldingAtHome/fah-cl ... t/pull/343 in the next client release.

You must be using the older v7 client. Try the newer v8 client which is available for Windows as well and is controlled via a web interface instead of a standalone GUI. It uses resource groups instead of slots which work in a different way and lets you do things like run two WUs on one GPU. https://...